BES-SVM: ENHANCED MACHINE LEARNING ALGORITHM FOR TELEGRAM SPAM CLASSIFICATION

Abstract

Support Vector Machine (SVM) has become a model of choice for intelligent spam detection due to its superiority compared to other machine learning classifiers. Despite this, the key to realizing its full capabilities lies critically in optimal hyperparameters tuning. To automatically tune the hyperparameters of SVM for improved spam detection on Telegram platform, this paper presents a hybrid approach, SVM optimized using Bald Eagle Search (BES) algorithm (BES-SVM). The BES algorithm, inspired by bald eagle foraging behavior, has gained attention for its simplicity and powerful global search capability. BES balances exploration and exploitation through selection, search, and swooping stages. Experiments were carried out in jupyter Notebook (Python 3) environment using dataset obtained from Kaggle. The proposed BES-SVM model achieved an overall detection accuracy of 95% in comparison to standard SVM (94%). Therefore, the BES-SVM framework provides an improved machine learning technique for smart Telegram spam detection.